Default empty DataFrame to dtype=object

[dalejung]

I ran into the following :

import pandas as pd

data = {}
columns = ['symbol', 'price']
df = pd.DataFrame(data, columns=columns)

df.symbol == 'AAPL'
# TypeError: Could not compare <type 'str'> type with Series

The problem being that an empty dataframe's dtype defaults to float. I've adjusted for it by checking the len of data and setting the dtype to object when at 0. But I feel like an empty DataFrame should act like an empty list, valid for all list operations. Which in lieu of #549, means casting the widest net possible with dtype=object.

[wesm]

Taking a look at this. I'm amazed (disturbed, a little?) so far how little code I'm having to touch (and how few tests are breaking as a result) to make this change

[dalejung]

<MrBurns>Excellent...</MrBurns>

I think the empty DataFrame Constructor is a corner case that short-circuits with a len check. So I can't imagine much depends on the float behavior.

[wesm]

Done. Hopefully won't wreak too much havoc